Documente Academic
Documente Profesional
Documente Cultură
Randomised Controlled
Trials (RCTs) – Sample Size:
The Magic Number?
Y H Chan
INTRODUCTION does not mean that that it is true (just that we do not
A common question posed to a biostatistician from have enough evidence to reject).
a medical researcher is “How many subjects do I We want to reject the null hypothesis but could
need to obtain a significant result for my study?”. be committing a Type I Error: rejecting the null
That magic number! In the manufacturing industry, hypothesis when it’s true. In a research study, there’s
it is permitted to test thousands of components in no such thing as “my results are correct” but rather
order to derive a conclusive result but in medical “how much error I am committing”. For example,
research, the sample size has to be “just large enough” if in the population, there are actually no differences
to provide a reliable answer to the research question. between two therapies (but we do not know, that’s
If the sample size is too small, it’s a waste of time why we are doing the study) and after conducting
doing the study as no conclusive results are likely the study, a significant difference was found which is
to be obtained and if the sample size is too large, given by p<0.05.
extra subjects may be given a therapy which perhaps There are only two reasons for this significant
could be proven to be non-efficacious with a smaller difference (assuming that we have controlled for
sample size(1). bias of any kind). One is, there’s actually a difference
Another major reason, besides the scientific between the two therapies and the other is by chance.
justification for doing a study, why a researcher wants The p-value gives us this “amount of chance”. If the
an estimate of the sample size is to calculate the cost of p-value is 0.03, then the significant difference due to
the study which will determine the feasibility of chance is 3%. If the p-value is very small, then this
conducting the study within budget. This magic difference happening by chance is “not possible” and
number will also help the researcher to estimate the thus should be due to the difference in therapies
length of his/her study – for example, the calculated (still with a small possibility of being “wrong”).
sample size may be 50 (a manageable number) but The other situation is not being able to reject
if the yearly accrual of subjects is 10 (assuming all the null hypothesis when it is actually false (Type II
subjects give consent to be in the study), it will take Error). As mentioned, the main aim of a clinical
at least five years to complete the study! In that case research is to reject the null hypothesis and we
a multicentre study is encouraged. could achieve this by controlling the type II error(2).
This is given by the Power of the study (1 – type II
STATISTICAL THEORY ON SAMPLE SIZE error): the probability of rejecting the null hypothesis
CALCULATIONS when it is false. Conventionally, the power is set
The Null Hypothesis is set up to be rejected. The at 80% or more, the higher the power, the bigger the
philosophical argument is: it is easier to prove a sample size required.
statement is false than to prove it’s true. For example, To be conservative, a two-sided test (more sample
Clinical Trials and we want to prove that “all cats are black”, and even size required) is usually carried out compared to a
Epidemiology if you point to me black cats everywhere, there’s one-sided test which has the assumption that the
Research Unit
226 Outram Road still doubt that a white cat could be lying under a test therapy will perform clinically better than the
Blk A #02-02
Singapore 169039 table somewhere. But once you bring me a white cat, standard or control therapy.
Y H Chan, PhD the hypothesis of ‘all cats are black’ is disqualified.
Head of Biostatistics Hence if we are interested to compare two SAMPLE SIZE CALCULATIONS
Correspondence to: therapies, the null hypothesis will be “there is no To estimate a sample size which will ethically
Y H Chan
Tel: (65) 6317 2121 difference” versus the Alternative Hypothesis of answer the research question of an RCT with a reliable
Fax: (65) 6317 2122 “there is a difference”. From the above philosophical conclusion, the following information should be
Email: chanyh@
cteru.gov.sg argument, not being able to reject the null hypothesis available.
Singapore Med J 2003 Vol 44(4) : 173
Type I error and Power(5) Table I shows the required sample size per group
The type I error is usually set at two-sided 5% and for π1 & π2 in steps of 0.1for powers of 80% & 90% at
power is at 80% or 90%. two-sided 5%.
Table I
π 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0.1 199 (266) 62 (82) 32 (42) 20 (26) 14 (17) 10 (12) 7 (9) 5 (6)
0.2 – 294 (392) 82 (109) 39 (52) 23 (30) 15 (19) 10 (13) 7 (9)
0.3 – 356 (477) 93 (125) 42 (56) 24 (31) 15 (19) 10 (12)
0.4 – 388 (519) 97 (130) 42 (56) 23 (30) 14 (17)
0.5 – 388 (519) 93 (125) 39 (52) 20 (26)
0.6 – 356 (477) 82 (109) 32 (42)
0.7 – 294 (392) 62 (82)
0.8 – 199 (266)
Numbers in ( ) are for 90% power
174 : 2003 Vol 44(4) Singapore Med J
Table II
δ
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
80% power 1,571 394 176 100 64 45 33 26 21
90% power 2,103 527 235 133 86 60 44 34 27
Table III
δ
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
80% power 792 200 90 52 34 24 19 15 12
90% power 1,052 265 119 68 44 32 24 19 15